Ultrasonic-assisted extraction for phenolic compounds and antioxidant activity of Moroccan Retama sphaerocarpa L. leaves: Simultaneous optimization by response surface methodology and characterization by HPLC/ESI-MS analysis

Abstract

This study was designed to optimize the ultrasound-assisted extraction of phenolic compounds and the antioxidant activity of Moroccan Retama sphaerocarpa extracts using response surface methodology (RSM). A central composite design has been conducted to investigate the effects of three factors: extraction period (X₁), solvent concentration (X₂), and solvent-to-material ratio (X₃) on extraction yield, total phenolic content (TPC), flavonoids content (TFC), and antioxidant activity. The obtained results showed that the experimental values agreed with the predicted ones, confirming the capacity of the used model for optimizing the extraction conditions. The best extraction conditions for the simultaneous optimization were an extraction time of 38 min, a solvent concentration of 58%, and a solvent-to-material ratio of 30 mL/g. Under these conditions, the optimized values of yield, TPC, TFC, and DPPH-radical scavenging activity (DPPH_IC50) were 18.91%, 154.09 mg GAE/g, 23.76 mg QE/g, and 122.47 μg/mL, respectively. The further HPLC/ESI-MS analysis of the obtained optimized extract revealed the presence of 14 phenolic compounds with piscidic acid, vitexin, and quinic acid as major compounds. These research findings indicate promising applications for efficiently extracting polyphenolic antioxidants, especially in the food industry.

Graphical abstract

Highlights

• •Phenolic compounds were obtained by ultrasound-assisted extraction from Moroccan Retama sphaerocarapa L. leaves.
• •Extraction yield, total phenols and flavonoids content and antioxidant activity were successfully optimized by response surface methodology.
• •Mathematical models were developed based on extraction period, temperature and solvent-to-sample ratio.
• •Extract obtained under optimal conditions was characterized by HPLC/ESI-MS.
• •Piscidic acid, vitexin, and quinic acid were major compounds in Retama sphaerocarapa L. leaves.

1. Introduction

Natural products discovery has recently gained considerable interest and has experienced a pronounced revival in the last two decades as the most successful class of drug leads [1]. They have received increasing attention especially with the worldwide pharmaceutical industry, with the growing population awareness and the high demand for herbal remedies promoting health advantages, cosmetics, and food processing industries [2]. Research on plant-based antioxidants is prompted by the fact that oxidative stress has a role in several illnesses [3].

Secondary metabolites which are often employed in medicines, nutraceuticals, food additives, and fine chemicals, are responsible for plants’ therapeutic effects [4]. Among these phytochemicals, plants synthesize various different phenolic compounds with different physicochemical characteristics that are beneficial for human health [5].

Extraction constitutes a crucial step in recovering phytochemicals from plant matrix and biomass [6]. This facilitates the emergence of more effective and sustainable extraction techniques for plant phytochemicals [2]. Different extraction methods have been extensively used to extract plant-derived phenolic compounds, such as maceration, supercritical fluid extraction, percolation, microwave-assisted extraction, and Soxhlet extraction [7]. However, they are still negatively described as time-consuming, requiring large amounts of solvent, and giving low extraction yields [8]. Among the unconventional methods, ultrasound-assisted extraction (UAE) is an innovative technique that has recently earned greater attention as a low-cost, simple, and effective substitute of conventional extraction techniques to improve the extraction of bioactive compounds [9]. UAE offers improved reproducibility, easier manipulation, less solvent usage, and lower energy input than other procedures [10]. This method employs sound waves, which facilitate solvent penetration into the sample matrix by expanding the contact surface between the solid and liquid phases. As a result, the solutes quickly diffuse from the solid to the solvent, increasing the extract yields [11].

Performing extractions at low temperatures is another advantage of UAE. Indeed, this may reduce heat losses caused by high temperatures and prevent the degradation of biologically active substances due to side undesirable reactions such as hydrolysis, ionization, and/or oxidation [12], which makes them undesirable from an economic perspective [13]. Several bioactive components, including polysaccharides, flavonoids, anthocyanins, and phenolic compounds, have been effectively extracted using the UAE [14]. Compared to conventional techniques such as maceration or Soxhlet extraction, UAE generally gives higher antioxidant yields and markedly reduces extraction time [15].

Retama sphaerocarpa is the plant investigated in the present study. It's a perennial leguminous shrub with evergreen photosynthetic stems that belongs to the Retama genus. This plant grows in the Mediterranean region of North East Africa, the Iberian Peninsula, and North Africa [16]. Retama species may flourish in various soil types and climatic circumstances since they can withstand arid conditions [17]. This plant has been traditionally used to cure various illnesses, including diabetes, rheumatism, and inflammations [18]. Fresh fruits are thus typically used to cure diarrhea, while flowers infusions are generally used to treat liver problems [19]. The plant has been also applied as a purgative, anthelmintic, healing agent after circumcision, vulnerary, antibacterial, and sedative for the treatment of local wounds, skin ulcers, and wounds [20]. The pharmacological activities of Retama species has been widely studied, revealing antioxidant [21], antibacterial [22], hepatoprotective [23], anti-inflammatory [24], anti-proliferative [25], and antiviral activities [26]. These beneficial effects are obviously due to the presence of various phytochemicals such as alkaloids, fatty acids, phenolic acids, flavonoids, and other antioxidants known for their ability to lower the risk of degenerative diseases related to oxidative stress [27]. Given the importance of the extraction process as the first step for the separation and isolation of bioactive compounds, we were interested to explore the phytochemical extraction of R. sphaerocarpa leaves using UAE. In order to enhance the recovery of target compounds, the extraction parameters will be optimized. Although several phytochemicals in Retama extracts have been previously investigated, no research has been done to determine the best conditions for extracting bioactive compounds from the leaves of this species. Moreover, and to the best of our knowledge, no research has been done on recovering R. sphaerocarpa’s phytochemicals from Morocco. Therefore, this work aimed to find the best operating conditions for extracting bioactive products from the leaves of this species using the response surface methodology (RSM) technique and further determine its phenolic profile. The studied responses were the extraction yield, total phenolic and flavonoid contents and antioxidant activity. The extraction effectiveness is affected by several variables including solvent type and concentration, solid-solvent ratio, time, temperature, frequency of sonication and particle size [28]. Thus, the optimization of UAE allows the testing of several parameters likely to influence the extraction of phenolic compounds. Based on literature data, the most studied parameters in UAE optimization are the extraction time, the material-to-solvent ratio and the solvent concentration [[28], [29], [30], [31], [32], [33], [34]]. From this background, these three parameters were chosen to study their effect on the three responses yield, TPC, TFC and DPPH_IC50.

The present research reported thus for the first time an optimization study for the recovery of phenolic compounds from R. sphaerocarpa to obtain extracts with better antioxidant properties. The chosen central composite design (CCD) allowed us to investigate the impact of extraction parameters and their interaction on the studied responses, and then identify the best-operating conditions leading to their optimization.

2. Materials and methods

2.1. Plant material

Retama sphaerocarpa aerial parts were harvested in April 2020 from Ouaouizerth in the Middle-High Atlas near Azilal, Morocco. The plant identification was achieved at the laboratory of plant biology and physiology at the Scientific Institute in Rabat. The aerial plant parts were washed, dried to a constant weight for seven days at room temperature in the shade, grounded using an electric grinder (Taurus, Barcelona, Spain), and sieved to obtain a fine powder (particle size 500 μm). The powders were packed in a fresh-keeping polyethylene bags and stored in a refrigerator at 4 °C until needed.

2.2. Sample preparation

The extraction process was performed using an ultrasonic bath JEULIN no. 701 340 (bath power 220 V and continuous mode at 40 kHz). In a 100 mL capped brown flask, 5 g of R. sphaerocarpa powder was mixed with a specified amount of the extraction solution ethanol: water at a given concentration. The combination was maintained in a water bath at 50 °C at the same distance from the ultrasound sides under prescriptive ultrasonic power for a specific time according to the experiment design. This temperature (50 °C) was chosen after a preliminary test carried out by fixing all the parameters and varying only the temperature from 30 °C to 60 °C. The suspensions were combined and filtered through Whatman's paper. The resulting filtrate was vacuum-vaporized at 40 °C on a rotary evaporator (Buchi R-210, Switzerland) and then refrigerated at 4 °C.

2.3. Determination of total phenolic content (TPC)

The Folin-Ciocalteu reagent was used to determine the total phenolic compounds concentration as described by Singleton et al. [35], with slight modifications. Briefly, 0.1 mL of each extract was combined with 0.1 mL of the Folin-Ciocalteu reagent for 5 min at room temperature. After neutralization with 10 mL of sodium carbonate solution 7.5%, the mixture was incubated for an additional hour. The absorbance was measures at 765 nm using a Lambda 25 Perkin-Elmer spectrophotometer (PerkinElmer, Waltham, MA, USA) having a wavelength range of 190–1100 nm and one fixed bandwidth equipped with 10 mm (path width) quartz cell. Gallic acid (0–0.225 g gallic acid/mL) was used as a standard to produce a linear calibration curve. The obtained results were expressed as mg gallic acid equivalents per gram of extract (mg GAE/g).

2.4. Determination of total flavonoid content (TFC)

Total flavonoid content was estimated by the colorimetric method using the aluminum chloride method previously reported by Lin et al., [36]. An aliquot of 500 μL of each extract was thoroughly mixed with 100 μL of 10% AlCl₃, 1.50 mL of 95% ethanol, 100 μL sodium acetate 1 M, and 2.80 mL distilled water. The absorbance was then determined at 415 nm. A quercetin standard calibration curve was used to carry out the quantification, and the findings were represented as mg of quercetin equivalents per gram of extract (mg QE/g).

2.5. DPPH free radical-scavenging activity

The capacity of R. sphaerocarpa to scavenge the DPPH radical was evaluated according to the method given by Pandey et al. [30]. Fifty μL of the explored samples at various concentrations were dissolved in methanol and then added to 2 mL of a 60 mM methanol solution of DPPH. The absorbance was recorded at 517 nm after 20 min at room temperature in the dark. Methanol with DPPH solution was used as a negative control, and the inhibition percentage of the DPPH free radical was calculated by using the following equation (Eq. (1)):

$$\%inhibition=\frac{\left(A_0-A_1\right)}{A_0}\times 100$$ (1)

A₀ stands for the sample used as a blank, and A₁ for the test sample's absorbance. By plotting the inhibition percentages against the sample concentrations, the sample concentration providing 50% inhibition (IC₅₀) was determined.

2.6. HPLC/ESI-MS analysis

The phenolic profile of the obtained extracts under optimized conditions has been explored with liquid chromatography coupled to a mass spectrometry detector. Analysis was performed on an Ultimate 3000 UHPLC apparatus consisting of a surveyor quaternary pump coupled to a PDA detector (200–600 nm), an electrospray ionization source, and an LCQ Advantage ion trap mass Thermo scientific spectrometer supplied by Orbitrap analyzer. A BDS Hypersil C18 column (150 mm × 4.6 mm × 5 μm) was used as a stationary phase. Water (A), acetonitrile (B), and acetonitrile (99:1, v/v), both containing 0.1% formic acid, formed the mobile phase at a flow rate of 0.48 mL/min. The used gradient was as follows: 0–5 min: 1% B, 5–8 min: 1–4.5% B, 8–20 min: 4.5–12% B, 20–22 min: 12–12.8% B, 22–27 min: 12.8–13.3% B, 27–33 min: 13.3–14.5% B, 33–48 min: 14.5–30% B, 48–55 min: 30–100% B, 55–59.5 min: 100–1% B then the column's re-equilibration for 5 min [27]. The column temperature was set at 45 °C, and the UV-visible detection range was set from 200 to 600 nm. UV spectra were also collected at 280 nm and 340 nm. Solutions with a concentration of 10 mg/mL were produced in MeOH: H₂O (50:50). A 0.2-m PTFE syringe filter was used to filter the resulting solution, and 20 L of the filtrate was then added to the HPLC apparatus. Phenolic compounds were identified based on UV and MS spectra in negative ion mode and by comparing the observed spectroscopic characteristics with those of literature data.

2.7. Experimental design

The optimal combination of extraction factors for the phenolic compounds from R. sphaerocarpa was determined using a central composite design (CCD). Three levels (lower (1), middle (0), and upper (+1)) of a three-factor method were used in the tests. As independent variables that should be adjusted for the extraction, three primary parameters impacting extraction efficiency, namely, extraction time (min, X₁), ethanol concentration (%, X₂), and solvent-to-material ratio (mL/g, X₃), were selected. Their uncoded and coded values are shown in Table 1. The experimental domains of each factor were enlarged to cover more fluctuations of the responses, and then, to recognize a more diverse set of effects. A central composite design (CCD) with 17 runs in random order and three replicates in the central point was employed to forecast the parameters' linear, quadratic, and interaction effects. The chosen CCD consists of factorial points (experiments 1–8), axial points with α = 1 (experiments 9–14), and center points (experiments 15–17).

Table 1.

Actual and coded levels of the independent variables used for the central composite design.

Independent variables	Coded variables	Variable levels			Unit
		−1	0	1
Extraction time	X1	20	40	60	Min
Solvent concentration	X2	40	60	80	% (v/v)
Solvent-to-material ratio	X3	10	20	30	mL/g

2.7.1. Fitted model and statistical analysis

Multiple regression analysis was used to fit the data to the quadratic polynomial model shown in equation (Eq. (2)):

$$Y=b_0+b_1{X}_1+b_2{X}_2+b_3{X}_3+b_{12}{X}_1{X}_2+b_{13}{X}_1{X}_3+b_{23}{X}_2{X}_3+b_{11}{X}_1^2+b_{22}{X}_2^2+b_{33}{X}_3^2+ϵ$$ (2)

where Y stands for the studied responses (extraction yield, total polyphenol content, flavonoid content, and DPPH free radical scavenging activity expressed by DPPH_IC50). b₀ represents the constant of the model, calculated as the average value of responses when all responses were at their level 0; b₁, b₂, and b₃ are coefficients of the main terms; b₁₁, b₂₂, and b₃₃ are the coefficients of quadratic terms; b₁₂, b₁₃, and b₂₃ are the coefficients of the interaction terms; X₁, X₂, and X₃ are the independent variables, while ε is the error term.

An analysis of variance (ANOVA) test at a 95% confidence level was performed to test the validity of the postulated model. Mean squares (MS) were obtained by dividing the sum of squares (SS) for each variation source (regression and residual) by the degree of freedom (DF). The F-value sufficiently describes the fluctuation of the data around its mean, and the model was statistically significant based on F_{ratio (R/r)}, which reflects the ratio between the mean square regression (R) and the mean square residual (r) [37].

The regression quality was measured using the coefficient of determination (R²), while the significance of the coefficients was examined based on their p-value using the t-student test. The coefficient is statistically significant if the p-value is less than 0.05 [37]. This analysis was conducted using SAS JMP® software v.14 and Expert Design software v.12. The comparison of means was performed using the ANOVA F-test followed by Tukey’s HSD test. This test compares each pair of means separately and reports on the pattern of differences between the means [38]. In the presentation of the results, the means considered statistically equivalent are attached with identical letters.

2.7.2. Optimization tools

The contour plot based on iso-response curves was employed to identify the areas of compromise leading to the desired response [39]. The optimal setting with a level of compromise was determined using the “Desirability” tool. With a percentage between 0 and 1, this tool allowed us to provide the optimal adjustment. When the components result in an undesirable reaction, the number 0 is assigned. In contrast, the maximum intended response is represented by the value 1, and each response was expressed as the mean ± standard deviation (SD). These techniques are highly helpful in determining the significance of the interactions between independent and dependent variables.

3. Results and discussion

3.1. Central composite design

The efficiency of phenolic compounds extraction, as well as the parameters that influence the antioxidant activity of R. sphaerocarpa, were studied by RSM. Table 2 lists the CCD parameters settings and the results of each experiment for yield, TPC, TFC, and DPPH_IC50 measurements. Every response averaged three repetitions, and the trials were conducted after randomization. It is worth noting that the recorded means for each of the four responses demonstrate a statistically significant difference (p-value < 0.05). Additionally, Tukey's test indicates the statistically identical means. This suggests that altering the operating circumstances significantly impacts the examined responses.

Table 2.

Matrix of Central composite design with actual, predicted and residual results for Yield, Total phenol contents, Total flavonoids content and DPPH_IC50.

N° of experiments*	Factor’s setting			Observed responses (**,***)
	Time (min)	Solvent Concentration (% v/v)	Solvent to material Ratio (mL/g)	Yield (%)			TPC (mg GAE/g)			TFC (mg QE/g)			DPPHIC50 (μg/mL)
				Actual	Predicted	Residual	Actual	Predicted	Residual	Actual	Predicted	Residual	Actual	Predicted	Residual
1	20	40	10	15.05 ± 0.16c	14.42	0.63	93.85 ± 0.23a	88.38	5.47	23.28 ± 1.19cde	23.07	0.21	197.20 ± 1.48i	199.91	−2.71
2	60	40	10	25 ± 0.20fg	24.23	0.77	123.2 ± 2.46de	128.74	−5.54	25.03 ± 0.47ef	22.62	2.41	134.5 ± 2.25c	133.21	1.29
3	20	80	10	11.66 ± 0.91b	11.59	0.07	110.45 ± 2.19c	115.94	−5.49	34.3 ± 3.22ij	32.24	2.06	188.85 ± 3.73h	187.77	1.08
4	60	80	10	24.59 ± 1.39fg	22.60	1.99	131.54 ± 0.64f	131.07	0.47	30.5 ± 2.12gh	32.97	−2.47	144.2 ± 4.27d	145.30	−1.10
5	20	40	30	15.5 ± 2.62c	17.13	−1.63	99.75 ± 2.40b	100.11	−0.36	19.47 ± 1.26abc	16.84	2.63	202.3 ± 2.83j	204.38	−2.08
6	60	40	30	26.5 ± 0.45g	26.20	0.30	144.02 ± 0.31h	138.43	5.59	24.64 ± 0.20def	26.54	−1.90	163.33 ± 2.36f	167.60	−4.27
7	20	80	30	12.43 ± 1.12b	12.83	−0.40	120.55 ± 4.89d	114.91	5.64	31.86 ± 0.45hi	34.12	−2.26	174.02 ± 4.24g	178.49	−4.47
8	60	80	30	22.85 ± 0.23ef	23.11	−0.26	122.63 ± 1.23de	128.00	−5.37	44.94 ± 2.45k	44.99	−0.05	165.47 ± 4.70f	165.94	−0.47
9	20	60	20	14.40 ± 1.67bc	17.86	−3.46	106.90 ± 3.20c	101.81	5.09	22.15 ± 4.13bcde	24.35	−2.20	160.05 ± 2.05ef	158.61	1.44
10	60	60	20	21.46 ± 1.67de	19.47	1.99	100.64 ± 2.49b	106.14	−5.50	28.83 ± 2.38gh	27.25	1.58	182.45 ± 2.35h	171.16	11.29
11	40	40	20	25.55 ± 1.48g	25.62	−0.07	124.95 ± 1.33de	130.11	−5.16	16.45 ± 0.66a	19.80	−3.35	163.00 ± 3.75f	155.23	7.77
12	40	80	20	21.25 ± 0.35de	22.65	−1.40	143.42 ± 3.25h	138.68	4.74	36.35 ± 1.63j	33.62	2.73	153.30 ± 1.17e	148.33	4.97
13	40	60	10	10.10 ± 1.70a	8.77	1.33	122.33 ± 1.53d	127.60	−5.27	16.94 ± 1.30ab	19.58	−2.64	171.66 ± 5.03g	163.48	8.18
14	40	60	30	16.00 ± 0.17c	18.81	−2.81	159.17 ± 4.34i	154.32	4.85	26.81 ± 2.38fg	24.79	2.02	128.41 ± 6.02a	123.86	4.55
15	40	60	20	19.25 ± 1.47d	18.79	0.46	136.29 ± 2.42g	130.57	5.72	23.18 ± 2.86cde	22.76	0.42	132.00 ± 0.7ab	143.76	−11.76
16	40	60	20	19.30 ± 1.13d	18.79	0.51	127.38 ± 2.47def	130.57	−3.19	21.72 ± 0.83bcd	22.76	−1.04	138.20 ± 5.23cd	143.76	−5.56
17	40	60	20	20.75 ± 2.18de	18.79	1.96	128.87 ± 3.89ef	130.57	−1.70	24.64 ± 2.09def	22.76	1.88	135.60 ± 0.85c	143.76	−8.16

*Experiments were carried out after randomization.

**Each response is the average of three replicates with standard error.

***Means with different letters were significantly different at the level of p < 0.05.

3.2. Statistical validation of the postulated model

According to the analysis of variance (Table 3), the main effect of the regression was significant for all responses since the probability of the risk significance p-value was less than 0.05 (0.007, 0.0044, 0.0054, and 0.0026 for yield, TPC, TFC, and DPPH_IC50, respectively). The calculation of the F_Ratio(R/r) for the four responses showed higher values than those the F_(0.05;9;7) at the 95% confidence level, which is equal to 3.67. Furthermore, the calculated F_{Ratio(Lof/Pe)} for the four responses revealed lower values than those of F_(0.05;5;2) at a 95% confidence level equal to 19.29. For the lack of fit test, the likelihood of a p-value larger than 0.05, the models are well adjusted to the observations.

Table 3.

Variance analysis for the fitted models.

Source of variation	DF	Yield				TPC				TFC				DPPHIC50
		SS	MS	F	p-value	SS	MS	F	p-value	SS	MS	F	p-value	SS	MS	F	p-value
R	9	389.9	43.3	7.6	0.007*	4400.9	489.0	8.9	0.0044*	788.5	87.6	8.3	0.0054*	8085.0	898.3	10.6	0.0026*
r	7	39.8	5.7			385.3	55.1			73.7	10.5			591.9	84.6
Lof	5	38.4	7.7	10.6	0.08	339.8	68.0	3.0	0.26	69.4	13.9	6.5	0.13	572.5	114.5	11.8	0.079
PE	2	1.5	0.7			45.6	22.8			4.3	2.1			19.4	9.7
Total	16	429.7				4786.3					862.2			8676.9
R²		0.91				0.92				0.91				0.93

DF: degrees of freedom; SS: sum of squares; MS: mean square; R: regression; r: residual; Lof: lack of fit; Pe: pure error; R²: coefficient of determination; *: statistically significant.

Also, the coefficients of determination were equal to 0.91, 0.92, 0.91, and 0.93 for yield, TPC, TFC, and DPPH_IC50, respectively. In fact, these values demonstrated a strong correlation between experimental results and model predictions, as expressed in Fig. 1.

Fig. 1.

The red lines represent the curve of actual values as a function of predicted ones for the four responses yield (a), TPC (b), TFC (c), and DPPH_IC50(d). The horizontal blue lines represent the average of the observed values for the four studied responses.

The obtained graphs (Fig. 1) showed that the curve of the actual values versus the predicted ones was perfectly straight for all the studied responses. Accordingly, the data approached a straight line, indicating that the anticipated data were in agreement with the experimental ones.

3.3. Study of factors’ effects and fitted models

The effects of all the investigated parameters, their t-student statistical values, and the observed probability (p-value) are depicted in Table 4. These coefficients were calculated based on the coded values of the parameters.

Table 4.

Estimated regression coefficients for all responses and their level of significance p-value.

Terms	Coefficients	Yield		TPC		TFC		DPPHIC50
		Estimation	p-value	Estimation	p-value	Estimation	p-value	Estimation	p-value
Constant	b0	18.78	<0.0001a	130.568	<0.0001a	22.76	<0.0001a	143.755	<0.0001a
Time	b1	0.804	0.3219	2.165	0.3868	1.45	0.2010	6.277	0.0677
Solvent concentration	b2	−1.482	0.0902	4.282	0.1107	6.91	0.0003a	−3.449	0.2743
Solvent-to-material ratio	b3	5.02	0.0003a	13.363	0.0007a	2.61	0.0386a	−19.812	0.0003a
Time * Time	b11	−0.12	0.9362	−26.589	0.0006a	3.04	0.1692	21.128	0.0071a
Solvent concentration * Solvent concentration	b22	5.34	0.0080a	3.826	0.4265	3.95	0.0866	8.028	0.1960
Solvent-to-material ratio * Solvent-to-material ratio	b33	−5.00	0.0110a	10.391	0.0556	−0.58	0.7796	−0.087	0.9881
Time * Solvent concentration	b12	−0.36	0.6782	−3.19	0.2632	2.03	0.1209	−3.436	0.3256
Time * Solvent-to-material ratio	b13	−0.18	0.8349	−0.51	0.8510	2.54	0.0626	7.479	0.0550
Solvent concentration * Solvent-to-material ratio	b23	0.3	0.7326	−6.30	0.0472a	0.30	0.8045	6.059	0.1046

^aStatistically significant at 95% confidence level.

For the response yield, the constant b₀, the linear effects of the ethanol concentration b₃, and the quadratic effects of b₂₂ and b₃₃ have a significant impact since their p-values were lower than 0.05. The relationship between extraction yield and variables was described using the second-order polynomial equation (Eq. (3)):

$${\hat{\mathrm{Y}}}_{Yield}=18.78{+5.02\mathrm{X}}_3+5.34{\mathrm{X}}_{22}^2-5{\mathrm{X}}_{33}^2$$ (3)

Concerning the response TPC, and as given in Table 4, the constant b₀, the linear coefficient b₃, the quadratic term b₁₁, and the negative interaction term b₂₃ were statistically significant (p-value < 0, 05). These results are in accordance with previously reported findings [34], who observed that the independent quadratic factors greatly impacted the recovery of phenolic compounds. The response TPC could be expressed by the following second-order polynomial equation (Eq. (4)):

$${\hat{\mathrm{Y}}}_{TPC}=130.56{+13.36\mathrm{X}}_3-26.589{\mathrm{X}}_{11}^2-6.30{\mathrm{X}}_2{\mathrm{X}}_3$$ (4)

For the response TFC, the linear components b₀, b₂, and b₃ have significant coefficients. Consequently, the fitted model was represented by the following equation:

$${\hat{\mathrm{Y}}}_{TFC}=22.76{+6.91\mathrm{X}}_2{+2.61\mathrm{X}}_3$$ (5)

The constant b₀, linear terms b₃, and quadratic term b₁₁ were the statistically significant coefficients concerning the DPPH_IC50 response. Therefore, the fitted model was represented by the following equation (Eq. (6)):

$${\hat{\mathrm{Y}}}_{{DPPH}_{IC50}}=143.75{-19.81\mathrm{X}}_3+21.12{\mathrm{X}}_{11}^2$$ (6)

As shown by the generated models, the four responses depend on linear terms and quadratic terms and not on interaction terms except for the negative interaction between solvent concentration and solvent/material ratio for the response TPC.

3.4. Optimization of processing parameters

In order to judge the postulated models' quality and achieve the best value for the four responses, the contour plot based on the isoresponse curves was used. A preliminary analysis reveals that optimizing the four responses necessitates maximizing the solvent-to-material ratio. Since the isoresponse profile is a 2D representation, we must fix one of the three factors and plot the profile against the other two. We eventually decided to set this component to its optimum value in order to plot the isoresponse surfaces as a function of the two other factors.

Fig. 2 shows the four contour plots generated by fixing the third parameter (solvent-to-material ratio) into its optimal value (1:30) and representing the four responses as a function of the two parameters, extraction time and solvent concentration.

Fig. 2.

Surface plots showing the optimal parameters setting leading to the best compromise zone of the four responses yield (a), TPC (b), TFC (c), and DPPH_IC50(d) obtained as a function of the two parameters time and solvent concentration by fixing the solvent-to-material ratio at 1:30.

3.4.1. Optimization of the yield of extraction

As shown in the isoresponse graph (Fig. 2a), it can be seen that a yield of around 26% (red zone) can be achieved by fixing the solvent-to-material ratio to its maximum value and ensuring an extraction time between 35 and 60 min with a solvent concentration ranging from 40 to 42%. Moreover, the desirability function shows that we have a 99% chance of reaching a value of 27.6% by keeping the solvent concentration at 40%, the liquid/solid ratio at 25 mL/g, and the time factor at a maximum of 60 min (Fig. 3a). Similar results were obtained using ethanol at concentrations close to 50% to achieve a maximum value under fixed UAE conditions [15,40,41]. Our optimized yield was higher than that of conventional extraction methods for other retama genus. For the species Retama raetam, Rejab et al. [42] have found 15.36% and 2.55% as extraction yield obtained by maceration with ethanol and hexane, respectively. Moreover, it was reported that maceration with methanol 70% gave a yield of 23.26% and 18.13% from the stems and seeds of Retama monosperma, respectively [43].

Fig. 3.

Desirability plots showing the precise process setting leading to the optimal value for yield (a), TPC (b), TFC (c), and DPPH_IC50(d), separately.

3.4.2. Optimization of total phenol content

As schemed in the contour plot (Fig. 2b), a value higher than 158 mg AGE/g (red zone) can be achieved by fixing the solvent-to-material ratio factor to its maximum value and by applying a time of extraction between 36 and 47 min and a solvent concentration between 40 and 45%. Besides, the desirability plot (Fig. 3b) shows that, with a 41 min extraction period, a solvent concentration of 40%, and a solvent-to-material ratio of 30 mL/g, we can generate 160 mg GAE/g DE with a desirability of 99%. Our obtained TPC is higher than the one obtained for R. sphaerocarpa by maceration with a mixture of methanol/water (80/20, v/v) followed by extraction with chloroform, ethyl acetate and n-butanol. In this study, the obtained TPCs were equal to 121.4 ± 0.3 GAE/g DE, 73.9 ± 0.3 GAE/g DE and 111.7 ± 0.1 GAE/g DE, respectively [43]. The same observation was made when comparing our results with obtained TPC by maceration with methanol:water (50/50 (v/v)) from R. sphaerocarpa seeds, which reported 125.8 ± 2.0 mg GAE/g as TPC [27]. Our findings are in agreement with those obtained by Milić et al. [44] who showed that the liquid/solid ratio had an influence on the extraction of total phenolic using UAE and reached the highest value (35.23 mg GAE/g DE) under the conditions of a liquid/solid ratio of 30%. In recent work, Ez zoubi et al. [34] reported that increasing the liquid/solid ratio up to 30 mL/g enhanced TPC recovery. Interestingly, a more excellent TPC value was achieved when the liquid-to-solid ratio was high due to the increased concentration gradient. This can be explained by the fact that the surface area of the substances in contact with the solvent increased, hence enhancing the solubility of phenolic compounds in plant cells [45]. In addition to the linear effect of the solvent-to-material ratio parameter, a negative interaction was observed with the solvent concentration parameter. This means that the effect of solvent-to-material ratio was not the same in all solvent concentrations. The same findings were made for the optimization of ultrasound assisted extraction of phenolic compounds from sparganii rhizoma, in which a significant effect of interaction of ethanol concentration and solvent-to-material ratio [33].

Moreover, the quadratic effect of time was crucial in improving extraction efficiency (p-value = 0.0006). Our study showed a gradual increase in TPC recovery with an extraction time fluctuation from 20 to 40 min. Further increases beyond 40 min led to its reduction. Initially, a prolonged extraction period enhanced extraction efficiency, but afterwards, there was a clear continuous fall in the response without any impact on extraction yield. This is most likely caused by polyphenol degradation and a decrease in the solvent's polarity as previously reported [46]. This behaviour was further confirmed by other researchers, indicating that TPC recovery does not require a more extended extraction period [47]. Furthermore, previously reported data proved that a better concentration of phenolic compounds could be obtained using ultrasonic extraction while extraction time could be decreased compared to extraction by maceration [48].

3.4.3. Optimization of flavonoid content

The contour plot (Fig. 2c) showed that the predicted flavonoid content was around 40 mg EQ/g with a process set comprising a maximization of the three studied parameters. Also, the desirability function (Fig. 3c) confirms with a compromise of 99% that setting parameters at their maximum level leads to obtaining a maximum flavonoid content equal to 45 mg QE/g DE. This TFC value remains relatively high with respect to previously reported result obtained from R. sphaerocarpa by maceration [43], and largely high compared to that recorded through maceration of R. raetam [42]. According to a similar study, choosing the appropriate ratio of ethanol to water efficiently improved the extraction of valuable compounds like flavonoids [49]. However, our findings differed from those reported by Sendi et al. [2] where the optimum conditions for flavonoid recovery were a solvent-to-liquid ratio of 90 mL/g, a solvent concentration of 50%, and an extraction time of 50 min. As a result, the extraction process of flavonoids would be more feasible and efficient in potential applications that consume less solvent.

3.4.4. Optimization of DPPH_IC50

As shown by the significance of the regression coefficients (Table 4), the solvent-to-material ratio was the most important factor impacting the DPPH radical scavenging activity, followed by the extraction time. First of all, increasing the solvent-to-material ratio toward its highest value (30 mL/g) enhanced the extraction performance. For this purpose, this factor will be fixed at its upper bound during the generation of the contour plot (Fig. 2d). This graph showed that a value of 121.7 g/mL (red zone) can be obtained by increasing the solvent-to-material ratio and ensuring a time processing between 27 and 39 min and a solvent concentration between 45 and 64%. These findings are confirmed by the desirability plot (Fig. 3d), in which the maximum value of the DPPH_IC50 response that can be achieved is equal to 121 μg/mL with a desirability of 99.1%. This value can be obtained with a time processing of 35 min, a liquid-solid ratio of 30 mL/g, and a solvent concentration of 56%. By comparing our optimized DPPH_IC50 with previous studies obtained by conventional methods, we found that it is very satisfactory compared to the results obtained by Soxhlet for the species R. sphaerocarpa (307.69 ± 1.33 μg/mL and 252.03 ± 3.38 μg/mL for grains and stems, respectively) [27]. In addition, Belmokhtar et al. [50] have reported a DPPH_IC50 equal to 3.15 ± 0.2 mg/mL, 1.51 ± 0.11 mg/mL, 2.88 ± 0.15 mg/mL and 3.87 ± 0.22 mg/mL for the maceration of R. monosperma stems in chloroform, ethyl acetate, n-butanol and methanol 70%, respectively. Although the solvent concentration had no effect on antioxidant activity in our case, it has been established that the solvent concentration has a positive effect on antiradical activity. The change in solvent polarity might justify this result as the ethanol concentration changes [51]. A recent study, Yu et al. [52] observed the same effect of ethanol concentration on antioxidant capacity of Cimicifuga dahurica (Turcz.) extracts. These researchers obtained an optimum IC₅₀ equal to 2.48 ± 0.1780 μg/mL according to the following conditions: ethanol concentration of 56.21%, liquid/solid ratio of 14.65:1, and extraction time of 124 min. Besides, the predicted optimal extraction time for the best phenolic recovery was 41 min; this time was significantly longer than that for the best antioxidant capacity (35 min). In other words, a prolonged extraction period likely alters the extraction of phenolic chemicals, leading to a decrease in antioxidant activity. The obtained results are in agreement with the previous data by Benarfa et al. [53] who reported that excessive time might cause undesirable changes in the extracted compounds due to overexposure to ultrasonic waves.

3.4.5. Simultaneous optimization of all responses

The simultaneous optimization of all responses relies on the desirability function tool, which allows the best fit of all responses to be sought with a level of compromise. Before proceeding with the optimization of the four responses, the priority of the responses must be specified, as the maximization of one response may imply decreases in the others since the behavior of the four responses differs with regard to the studied parameters. Thus, the priority is given to the two responses, TPC and DPPH_IC50, and the optimization will be based on maximizing the first and minimizing the second.

The desirability plots in Fig. 4 indicated that optimizing all responses was possible with a desirability of 99% by ensuring an extraction time of 38 min, a solvent-to-material ratio of 30 mL/g, and a solvent concentration of 58%. The corresponding optimal response values for this setting are equal to 18.91%, a phenol content of about 154.09 mg GAE/g, a flavonoid content of approximately 23.76%, and a DPPH_IC50 of 122.47 μg/mL.

Fig. 4.

Desirability plots of simultaneous optimization showing the precise values of time, solvent concentration, and solvent-to-material ratio leading to the optimal values of yield, TPC, TFC, and DPPH_IC50.

3.4.6. Experimental validation of the optimal conditions

The accuracy of the generated models for the four responses extraction yield, TPC, TFC, and DPPH_IC50, was verified by a validation test. The test entails comparing this point's predicted results with those experimentally attained. The test coordinates were the optimal extraction conditions that gave the optimal values, producing the best ones. According to Table 5, the obtained experimental values belong to the confidence interval of the predicted values. Therefore, no noticeable difference can be observed between the experimental and predicted responses.

Table 5.

Predicted and experimental values for the test point realized by the optimal operating conditions.

		Yield (%)		TPC (mg GAE/g)		TFC (mg QE/g)		DPPHIC50 (µg/mL)
Parameters	Operating setting	Predicteda	Experimentalb	Predicteda	Experimentalb	Predicteda	Experimentalb	Predicteda	Experimentalb
Extraction time	38	18.91 ± 2.38	16.62 ± 1.23	154.09 ± 7.41	160.15 ± 3.2	23.76 ± 3.24	26.91 ± 1.12	122.47 ± 9.19	128.45 ± 2.12
Solvent-to-material ratio	30
Solvent concentration	58

^aThe predicted value is given with the standard deviation of the response calculated from the model.

^bThe observed value is the average of three replicates with standard error.

3.5. HPLC-ESI-MS analysis

The phenolic composition of R. sphaerocarpa extract was explored by liquid chromatography coupled to both PDA and MS detectors. As discussed below, fourteen compounds were identified, including two phenolic acids and twelve flavonoids. An example of the obtained chromatographic profile is indicated in Fig. 5. The obtained chromatographic and spectroscopic characteristics of each detected compound are gathered in Table 6.

Fig. 5.

LC/ESI-MS chromatographic profile recorded in the negative ion mode of R. sphaerocarpa extract obtained in the optimized extraction conditions.

Table 6.

Phenolic compounds profile in R. sphaerocarpa obtained by optimal conditions using UHPLC-ESI-MS analysis in negative ionization mode.

Peak N°	Compound	Rt (min)	UV (nm)	[M − H]-(m/z)	MS Product ions (m/z)	References	Percentage (%)
1	Quinic acid	3.75	225	191	173, 127, 111	[54,74]	3.86
2	Not identified	4.16	295	273	159	–	3.15
3	Piscidic acid	14.07	224, 275	255	193, 165, 119	[56,74]	21.02
4	Not identified	21.87	242, 373	370	239	–	1.04
5	Vitexin O-hexoside	26.49	232, 321	593	503, 473, 431, 311	[27,57]	11.35
6	Kaempferol di-O-glucoside	30.97	230, 348	609	447, 489, 285	[69,71]	4.86
7	Vitexin isomer	31.46	261, 340	431	353, 311, 283	[27]	27.60
8	Genistein O-hexoside	35.59	235, 264, 332	431	341, 311, 268	[65,72]	1.35
9	Vitexin O-hexoside	39.17	235, 266, 323	593	473, 431, 269, 311	[57]	3.82
10	Luteolin O-pentoside-O-glucoronide	40.27	233, 344	593	417, 285, 267, 241	[27]	1.73
11	Apigenin 6,8-di- C-hexoside	42.26	234, 262, 330	593	473, 353	[64,77,78]	3.32
12	Vitexin-3-hydroxy-3-methylglutaroyl	42.95	234, 320	575	473, 431, 311	[27]	2.13
13	Kaempferol O-rhamnoside	43.68	233, 269, 323	431	311, 285	[71]	3.55
14	Vitexin O-hexoside	45.39	233, 269, 316	593	473, 431, 311	[57]	4.17
15	Calycosin O-hexoside	46.74	233, 261	491	445, 283	[72,74,79]	3.86
16	Chrysin O-hexoside	47.87	233, 310	415	253	[73]	2.11

3.5.1. Phenolic acids

Compound 1 (Rt = 3.75) displayed a signal at m/z 191 corresponding to the deprotonated molecular ion [M−H]⁻. Other additional significant ion signals were located at m/z 173, 127, and 111 corresponding to [M−H–CO–2H₂O]⁻, [M−H–H₂O]⁻ and [M−H–2H₂O–CO₂]⁻ ions respectively in agreement with previously reported data of quinic acid [54]. This compound is a natural antioxidant compound widely distributed in Retama species. Quinic acid derivatives with caffeoyl groups were reported to have higher antioxidant activities than the free acid form [55].

Compound 3 (Rt = 14.07) exhibits an [M−H]⁻ ion signal at m/z 255. This compound was tentatively identified as piscidic acid on the basis of the fragment signals observed at m/z 193 corresponding to the [M−H–CHO₂–OH]⁻ ion. This was also confirmed by the presence of other signals at m/z 165 and 119 in agreement with previously reported data of piscidic acid [56].

3.5.2. Flavonoids

Compound 5 at Rt 26.49 min showed a peak signal at m/z 593 corresponding to the deprotonated molecule ion [M − H]⁻. Another signal was observed at m/z 431 and was associated with the loss of the O-hexoside moiety (−162 Da). The consecutive losses of C-hexosyl moiety (−90 and −120 Da) were observed in MS spectra with ion products characteristic of C-glycosyl flavones at m/z 503 and m/z 473, respectively [57]. Consequently, this compound was identified as vitexin-O-hexoside. Similarly, compounds 9 and 14 (Rt = 39.17 min and 45.39 min, respectively) were assigned to vitexin-O-hexoside isomers based on their UV properties characteristics of flavones with two typical absorption (one close to 320 nm and another around 260–270 nm). The proposed structures agreed with the obtained molecular ion signals observed at m/z 593 for both compounds. This was also reinforced by the obtained fragmentation pattern similar to apigenin-di-O-hexoside and especially with the characteristic fragmentations of O-, C-glycosyl flavones, with the losses of a C-hexosyl moiety 473 [M–H–120]⁻ and an O-hexoside moiety (−162 Da) [27]. Other product ions were observed at m/z 431 and m/z 311 caused by the loss of a CO group typical of such compounds [58]. As for compound 7, the signal at m/z 311 [MH-120]⁻ showed the existence of a C-glucosyl unit based on its [M − H]⁻ ion at m/z 431 pattern characteristic of a mono-C-glycoside [59]. Another vitexin derivative was detected at Rt = 42.95 (compound 12) and was tentatively identified as vitexin-3-hydroxy-3-methylglutaroyl. Its obtained UV spectrum was comparable to those of the mentioned above vitexin isomers. The MS spectra displayed a deprotonated molecular ion at m/z 575 along with a product ion at m/z 431 and an additional loss of 144 Da, showing the 3-hydroxy-3-methylglutaroyl moiety [60].

Compound 10 (Rt = 40.27 min) was concluded to be a flavone derivative based on its UV spectral characteristics. Its MS spectra displayed a [M − H]^- ion at m/z 593 that produced an ion at 417 [M-H-176]^- as a consequence of the neutral loss of 176 Da in agreement with the structure of glucuronide acid [61]. An additional signal was observed at m/z 285 consistent with luteolin aglycone moiety [62]. The MS spectra of this compound also revealed a pentosyl unit loss and further water loss typical of O-hexoside units [63]. The obtained data were in agreement with a luteolin O-pentoside-O-glucuronide structure [63].

Compound 11 (Rt = 42.26 min) was determined as apigenin-6,8-di C-glycoside (vicenin 2) based on its [M − H]^- at m/z 593. Typical fragment ion signals were observed at m/z 353 and 473 [M–H–120]^- representing sugar cross ring cleavages in agreement with the fragmentations observed for di C-glycosylflavone derivatives [64]. This was also confirmed through UV spectra which were similar to those of apigenin. The MS spectra of compound 8 presented a molecular ion signal at m/z 431 and a fragmentation pattern signals consistent with genistein hexoside product [65]. This is consistent with previously reported data on R. sphaerocarpa, revealing the presence of genistein derivatives in R. sphaerocarpa [66]. Such derivatives were reported as the most abundant components in Retama species [67] and are known to exhibit high antioxidant power [68].

Compound 6 (Rt = 30.97 min) showed a signal at m/z 609 corresponding to the deprotonated molecular ion. Other fragment signals were observed at m/z 447 [M–H–162]^– and m/z 285, resulting from the consecutive loss of two O-hexoside units and showing a typical fragmentations of a kaempferol aglycone moiety [69,70]. The kaempferol skeleton was confirmed based on the obtained ESI-MS spectral characteristics.

Compound 13 (Rt = 43.68 min) gave a [M − H]^– ion at m/z 431 and its fragment signal at m/z 285 [M–H–146]^– in agreement with a deprotonated kaempferol unit released after the neutral loss of rhamnosyl moiety (146 Da) [71]. Consequently, compound 13 was identified as kaempferol O-rhamnoside.

Compound 15 (Rt = 46.74 min) was tentatively identified as a methylated isoflavone derivative. Its MS spectra showed a deprotonated molecular ion [M−H]⁻ signal at m/z 490. Further ion signal was observed at m/z 283 [M–H–162] upon fragmentation as a result of the neutral loss of the O-hexoside unit. Since calycosin-O-hexoside presents the same UV and the same fragmentation pattern [72], compound 15 was identified as calycosin O–hexoside.

Compound 16 (Rt = 47.87 min) was identified as chrysin O-hexoside based on its [M − H]^– ion at m/z 415 and subsequent fragmentation route. The MS data of the deprotonated ion produced at m/z 253 suggests an O-glycoside substituted unit [73].

3.5.3. Relative quantitative analysis

After having studied the qualitative phytochemical composition of the obtained extract, the relative quantitative analysis of the individual identified phenolic compounds was investigated. The phenolic profile was explored using UHPLC chromatography and the obtained results are gathered in Table 6. The relative quantitation of the detected compounds has been made on the basis of the area % calculation procedure which reports the area of each peak in the chromatogram as a percentage of the total area of all peaks. This method supposes that all components respond equally in the detector and are all eluted from the column. Even if these criteria are may be not assured, the used method provides at least a suitable approximation of the relative amounts of the detected compounds.

The most prevalent classes of phenolic compounds detected in R. sphaerocarpa aerial parts were flavones and flavonols. Vitexin was the main compound found representing 27.60% of the extract followed by piscidic acid (21.02%) confirming the presence of phenolic acids in Retama species [74].

The examined sample also contained other flavonoids glycosides such as kaempferol di-O-glucoside (4.86%), genistein-O-hexoside (1.35%) and vitexin-O-hexoside (3.82%). Interestingly, the obtained data showed similar flavonoid content in the water: ethanol extracts both in terms of abundance and number of detected compounds compared to those previously detected in the methanol: water extracts [27]. One of the possible reasons for this similarity might be due to the solubility of R. sphaerocarpa leaves compounds in aqueous alcoholic solutions. Therefore, flavonoids can be best extracted by hydro-alcoholic mixture than pure alcohol or water [75]. The results obtained in the current study representing the first phytochemical investigation of R. sphaerocarpa leaves from Morocco are consistent with its previously reported phenolic composition [66,76]. From a qualitative point of view, the identified compounds have been previously reported in R. sphaerocarpa [27] with different percentages. The observed differences might be due to various variables including environmental, extraction or quantification techniques.

4. Conclusion

In this paper, a central composite design and UAE were successfully employed to optimize the extraction process of phenolic compounds from Moroccan R. sphaerocarapa. According to the obtained results, the simultaneous optimal extraction conditions were 38 min, 58%, and 30 mL/g for extraction time, solvent concentration, and solvent-to-material ratio, respectively. The experimental results agreed with the predicted values and provided a concept for the possible application of the obtained extracts in both scientific and industrial fields. Furthermore, the proposed UAE approach can offer an effective method for the simultaneous extraction of phenolic compounds and antiradical capacities from R. sphaerocarpa. Consequently, this optimized UAE method has demonstrated a potential application and provided theoretical guidance for discovering and refining potential antioxidants from Retama species. Moreover, the phenolic profile of R. sphaerocarapa proved its strong potential to serve as a source of phytochemical compounds that can be applied as bioactive agents. As a further perspective, it will be interesting to investigate other parameters influencing the extraction process, such as temperature and sonication frequency, and the mixture of solvents, which are relevant on an industrial scale and require more research and improvement.

Author contribution statement

Aafaf El Baakili, Mouhcine Fadil: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Nour Eddine ES-SAFI: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper.

Data availability statement

Data included in article/supp. material/referenced in article.

Funding

This work was supported by ANPMA/UM5R/CNRST (Project number PMA2019-4).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.